> 
> I was wondering why my PS/2 keyboard is always auto-detected (i386,
> 1.5.1), even when it is not present ?

yes. there is no checking if keyboard is actually present. i don't think
it's even possible.

> To be able to use either a PS/2 or an USB keyboard, I set up my kernel
> like this :
> 
> pckbc0		at isa?
> pckbd*		at pckbc?
> wskbd1 		at pckbd? console ?
> 
> ukbd*	at uhub? port ? configuration ? interface ?
> wskbd0	at ukbd? console ? mux 1
> 
> And here is the dmesg when I only have the USB keyboard plugged-in:


but you still could plug PS/2 keyboard while system is working
 
> [...]
> pckbc0 at isa0 port 0x60-0x64
> pckbd0 at pckbc0 (kbd slot)
> pckbc0: using irq 1 for kbd slot
> wskbd1 at pckbd0: console keyboard, using wsdisplay0
> [...]
> ukbd0 at uhub0 port 2 configuration 1 interface 0
> ukbd0: Sun Microsystems Type 6 USB, rev 1.00/1.02, addr 3, iclass 3/1
> wskbd0 at ukbd0 mux 1
> 
> This causes (minor) problems with keys like "caps lock" or "num lock"
> that try to send commands to the (unplugged) PS/2 keyboard, ending up in
> a "pckbc: cmd failed" for the PS/2 keyboard.

try removing pckbd from your kernel config.

> I think a basic test at boot time should detect that the keyboard is not
> present and just forget about it. Maybe I could try to send-pr a patch ?
>